The rise of AI in healthcare education: DeepSeek and GPT-4o take on the 2024 Taiwan Pharmacist Exam

Sep 29, 2026

Letter to the editor describing how large language models are reshaping medicine and pharmacy education, with focus on comparing open-source DeepSeek and GPT-4o for the 2024 Taiwan national pharmacist licensing examination. The letter outlines DeepSeek’s claimed strengths (multilingual processing, cost-effectiveness, openness) and discusses limitations in medical-domain use including annotated-data quality, verbose reasoning, ethics, regulation, data privacy, and the need for clinical validation. It reports a structured question sampling approach and records model responses for subsequent analysis.

LanguageEnglish

CategoryExam

FormatPDF

Pages2

Reading Time5 min

AccessFree to download

The rise of AI in healthcare education

Background: LLMs in medicine and pharmacy

DeepSeek overview and limitations in healthcare use

Rationale for testing in Taiwan pharmacist examinations

Experimental approach: question sampling and response recording

J Chin MedAssoc Letter to the editor The rise of AI in healthcare education: DeepSeek and GPT-4o take on the 2024 Taiwan Pharmacist Exam Ying-Mei Wanga,b,c, Tzeng-Ji Chend,e,f,* aDepartment of Medical Education and Research, Taipei Veterans General Hospital Hsinchu Branch, Hsinchu, Taiwan, ROC; bDepartment of Pharmacy, Taipei Veterans General Hospital Hsinchu Branch, Hsinchu, Taiwan, ROC; cSchool of Medicine, National Tsing Hua University, Hsinchu, Taiwan, ROC; dDepartment of Family Medicine, Taipei Veterans General Hospital Hsinchu Branch, Hsinchu County, Taiwan, ROC; eDepartment of Family Medicine, Taipei Veterans General Hospital, Taipei, Taiwan, ROC; fDepartment of Post-Baccalaureate Medicine, National Chung Hsing University, Taichung, Taiwan, ROC DEAR EDITOR, In recent years, the rapid development of large language model (LLM) technology has not only changed the landscape of artificial intelligence (AI) research but has also begun to have a profound impact on the fields of medicine and pharmacy. In the past, LLMs were primarily dominated by technology giants operating under closed commercial models, such as OpenAI’s GPT series and Google DeepMind’s Gemini. However, the rise of open-source LLMs is reshaping this competitive landscape and opening up new opportunities for the research community and industry. DeepSeek is one of the most popular open-source LLMs in recent times, not only for its strong technical performance but also for its strengths in multilingual processing, costeffectiveness, and openness.1 This model has made waves in the LLM market with its groundbreaking Mixture-of-Experts (MoE) architecture and reinforcement learning techniques, enabling it to compete with OpenAI’s GPT-4o despite limited computational resources, and even outperform it in mathematics, programming, and reasoning capabilities.2 Furthermore, DeepSeek’s open-source strategy and lower operating costs have quickly drawn global attention, attracting numerous scientists to explore its applications in various fields.3,4 This technological evolution is poised to drive considerable advances in both academia and industry. DeepSeek has demonstrated potential in scientific areas such as bioinformatics, computational chemistry, and neuroscience, achieving performance comparable to OpenAI’s GPT-4o in certain data analysis and diagnostic tasks.5 However, its application in the medical domain remains limited by factors such as the quality of annotated data, excessive transparency of reasoning leading to verbose output, and ethical and regulatory challenges, *Address correspondence. Dr. Tzeng-Ji Chen, Department of Family Medicine, Taipei Veterans General Hospital Hsinchu Branch, 81, Section 1, Zhongfeng Road, Zhudong, Hsinchu 310, Taiwan, ROC. E-mail address: [tjchen@vhct.gov.tw](tjchen@vhct.gov.tw)[ ](tjchen@vhct.gov.tw)(T.-J. Chen). Conflicts of interest: Dr. Tzeng-Ji Chen, an editorial board member at Journal of the Chinese Medical Association, had no role in the peer review process of or decision to publish this article. The authors declare that they have no conflicts of interest related to the subject matter or materials discussed in this article. Journal of Chinese Medical Association. (2025) 88: 338-339. doi: 10. 1097/JCMA.0000000000001220 Copyright © 2025, the Chinese Medical Association. This is an open access article under the CC BY-NC-ND license ([http://creativecommons.org/licenses/](http://creativecommons.org/licenses/)[ ](http://creativecommons.org/licenses/)[by-nc-nd/4.0/](by-nc-nd/4.0/)) including data privacy and liability issues.1 Additionally, although DeepSeek is freely accessible, its reliability in clinical decision-making requires further validation. Moreover, it still exhibits errors in some basic tasks, such as simple calculations.6 Comparing the applications of DeepSeek and ChatGPT in clinical research, both models can accelerate scientific research, streamline clinical data analysis and diagnostics,

FAQs

Why compare DeepSeek and GPT-4o in pharmacist examination contexts?

Because LLMs are increasingly used in medical research and education, but performance—especially for non-English tasks and clinical decision-related reliability—needs direct evaluation on professional exam content.

What strengths of DeepSeek are highlighted in the letter?

It is described as an open-source model with strong multilingual processing, cost-effectiveness, openness, and competitive performance on reasoning, mathematics, programming, and some scientific analysis tasks.

How were examination questions selected for the test?

A 25% random sample of questions was taken for each subject; when a subject contained two subfields, separate 25% samples were taken for each subfield, then entered directly into DeepSeek and ChatGPT and responses were recorded.

Sep 29, 2026

Upload by Theodore